Sub-field normalization in the multiplicative case: High- and low-impact citation indicators

This article uses high- and low-impact citation indicators for the evaluation of the citation performance of research units at different aggregate levels using a dataset of about 3.6 million articles published in 1998-2002 in the natural and the social sciences with a 5-year citation window. The difficulty is that a large proportion of individual articles are assigned to multiple subfields. To control for wide differences in citation practices at the subfield level, we apply a novel normalization procedure in the multiplicative approach in which each paper is wholly counted as many times as necessary in the several categories to which it is assigned at each aggregation level. The methodology is applied to a partition of the world into three geographical areas: the USA, the European Union (EU), and the Rest of the World. The main findings are the following two. (1) Although normalization does not systematically bias the results against any area, it reduces the US/EU high-impact gap in the all-sciences case by a non-negligible 14.4%. (2) The dominance of the USA over the EU in the basic and applied research published in the periodical literature is almost universal at all aggregation levels. From the high-impact perspective, for example, the USA is ahead of the EU in 77 out of 80 disciplines, and all of 20 fields. For all sciences as a whole, the US high-impact indicator is 61% greater than that of the EU.